J OURNAL OF Journal of Petrology, 2017, Vol. 58, No. 1, 85–114 doi: 10.1093/petrology/egx006 P ETROLOGY Advance Access Publication Date: 15 March 2017 Original Article
Pleistocene to Holocene Growth of a Large Upper Crustal Rhyolitic Magma Reservoir beneath the Active Laguna del Maule Volcanic Field, Central Chile Nathan L. Andersen1*, Brad S. Singer1, Brian R. Jicha1, Brian L. Beard1, Clark M. Johnson1 and Joseph M. Licciardi2
1Department of Geoscience, University of Wisconsin–Madison, Madison, WI 53706, USA; 2Department of Earth Sciences, University of New Hampshire, Durham, NH 03824, USA
*Corresponding author. E-mail: [email protected]
Received June 17, 2016; Accepted January 26, 2017
ABSTRACT The rear-arc Laguna del Maule volcanic field (LdM) in the Andean Southern Volcanic Zone, 36 S, is among the most active latest Pleistocene–Holocene rhyolitic centers globally and has been inflating at a rate of > 20 cm a–1 since 2007. At least 50 eruptions during the last 26 kyr allow for a thorough interrogation of changes in the physical and chemical state of this large, 20 km diameter, silicic system. Trace element concentrations and Sr, Pb and Th isotope ratios indicate that the mafic pre- cursors to the LdM rhyolites result from mixing between partial melts of garnet-bearing mantle and crust in Th-excess and partial melts of garnet-free crust in U-excess. The 238U/230Th ratios of the LdM lavas are decoupled from the slab fluid signature, similar to several recently studied fron- tal arc volcanic centers in the Southern Volcanic Zone. A narrow range of radiogenic isotope com- positions and increasing isotopic homogeneity with differentiation indicate that silicic magma is generated by magma hybridization and crystallization in the upper crust with limited involvement of older, radiogenic material. New 40Ar/39Ar and 36Cl ages reveal a wide footprint of silicic volcan- ism during the early post-glacial (25–19 ka) and Holocene (c. 8–2 ka) periods, but focused within a single eruptive center during the interim period. Subtle temporal variations in trace element com- positions and two-oxide temperatures indicate that these eruptions, issued from vents distributed within a similar area, tapped at least two physically discrete rhyolite reservoirs. This compositional distinction favors punctuated extraction and ephemeral storage of the erupted magma batches. Frequent mafic recharge incubates this long-lived, growing shallow silicic magma reservoir above the granite eutectic, which favors magma interactions over rejuvenation of near- to sub-solidus silicic cumulates. A long-term rate of mass addition—extrapolated from surface deformation accu- mulated over the past decade—is comparable with those that have produced moderate- to large- volume caldera-forming eruptions elsewhere.
Key words: rhyolite; Andes Southern Volcanic Zone; magma chamber; geochronology; radiogenic isotopes
INTRODUCTION (ignimbrite) deposits are often interpreted to reflect the Large silicic volcanic systems are of great interest because structure of the pre-eruption magma reservoir (e.g. they generate caldera-forming eruptions that disperse Hildreth, 1981). The composition and ages of major and enormous quantities of ash over a vast area. accessory phases can provide records of magma accumu- Heterogeneities in the resulting pyroclastic fall and flow lation, crystallization, and mixing on both short (100–102
VC The Author 2017. Published by Oxford University Press. All rights reserved. For permissions, please e-mail: [email protected] 85 86 Journal of Petrology, 2017, Vol. 58, No. 1 year) and long (104–106 year) timescales (e.g. Vazquez & melting of the deep crust (up to 70% depending on the Reid, 2004; Charlier et al., 2005, 2008; Wark et al., 2007; magma flux and lithology of the crust), fractional crys- Costa, 2008; Reid, 2008; Reid et al., 2011; Wotzlaw et al., tallization of hydrous basalt, and mixing of the resulting 2013, 2015; Chamberlain et al., 2014a, 2014b). differentiates and crustal melts. Shallow systems are Complementing these records are studies of smaller pre- assembled incrementally from these lower crustal ‘hot and post-caldera silicic eruptions that record the longer zones’ (Annen et al., 2006), but undergo limited chem- thermochemical context that produced the caldera- ical differentiation following shallow magma emplace- forming system, particularly when the earlier or subse- ment. Thus, the volume of eruptible magma is primarily quently erupted material is physically distinct from the a function of the magma flux to the upper crust (e.g. caldera-forming system or the caldera-collapse event pro- Glazner et al., 2004; Annen et al., 2006; Annen, 2009; duces a structural realignment of the shallow magma sys- Gelman et al., 2013). tem (Metz & Mahood, 1985, 1991; Sutton et al., 2000; The investigation of pre-caldera silicic eruptions can Charlier et al., 2005; Smith et al., 2005, 2010; Simon et al., provide clues to the physical and thermal evolution that 2007; Wilson & Charlier, 2009; Bachmann et al., 2012; sets the stage for the assembly and eruption of a volu- Barker et al., 2015). minous silicic magma reservoir. Pre-caldera eruptive re- The archetype model of voluminous silicic magma cords can be limited owing to infrequent eruptions, systems involves crystallization of mafic to intermediate poorly resolved geochronology, burial or destruction by forerunners in the middle to upper crust, yielding an subsequent caldera-forming events (Metz & Mahood, intermediate to silicic crystal mush—an extensive 1991; Stix & Gorton, 1993; Wilson et al., 2009). crystal-rich (>60% solid) reservoir containing evolved Nevertheless, such records have proven useful in iden- interstitial melt. Crystal-poor eruptible magma bodies tifying changes in the mafic flux to the upper crust, the are assembled by progressive extraction and accumula- amalgamation of previously discrete magma reservoirs, tion of melt from these crystal-rich domains (Bachmann and placing limits on the longevity of the subsequent & Bergantz, 2004; Hildreth, 2004) or remelting of silicic caldera-forming reservoir (e.g. Metz & Mahood, 1991; cumulate during magma recharge events (Mahood, Simon et al., 2007; Bindeman et al., 2008; Wilson & 1990; Wolff et al., 2015; Evans et al., 2016). The relative Charlier, 2009; Chamberlain et al., 2014b). importance of these mechanisms varies between Understanding the recent magmatism at historically ac- caldera-forming systems as well as within zoned ignim- tive rhyolitic volcanic centers (e.g. Miller, 1985; Hildreth, brites produced during individual events (e.g. Vazquez 2004; Smith et al., 2005; Castro & Dingwell, 2009; & Reid, 2004; Charlier et al., 2005; Bindeman et al., Hildreth & Fierstein, 2012; Rawson et al., 2015) allows 2008; Wotzlaw et al., 2013, 2015; Chamberlain et al., for the interrogation of the structure of the magma res- 2014a, 2014b; Evans et al., 2016). ervoir, the petrogenesis of rhyolites, the physical and Departures from the model of progressive rhyolite thermal processes preceding the recent eruptions, and extraction have been noted at large silicic systems such their evolution through time. Such systems are poten- as Taupo Volcano and Yellowstone involving a greater tial sites of caldera-forming eruptions and, taken to- proportion of remelting of silicic forerunners and the gether, this information is valuable in evaluating the amalgamation of distinct rhyolite melts, potentially cat- possible style of future eruptions and establishing a alyzed by extensional tectonics (Smith et al., 2004, 2010; context in which to better interpret seismic, magnetotel- Charlier et al., 2005, 2008; Wilson et al., 2006; Shane luric, geodetic, and gravity observations (e.g. Singer et al., 2007, 2008; Bindeman et al., 2008; Wilson & et al., 2014). Charlier, 2009; Allan et al., 2013; Be´ gue´ et al., 2014; The rear-arc Laguna del Maule (LdM) volcanic field Storm et al., 2014). Brief repose periods following the (Fig. 1) produced two dacitic to rhyodacitic caldera- eruption of compositionally distinct pre-caldera rhyo- forming eruptions during the mid-Pleistocene. A recent lites, durations of zircon crystallization, and crystal resi- concentration of silicic volcanism has yielded at least 50 dence based on solid-state diffusion kinetics indicate rhyolitic eruptions in the last 26 kyr; thus LdM is among that the assembly of 102–103 km3 eruptible rhyolite the most frequently erupting active rhyolitic volcanic magma bodies in these systems occurred more rapidly centers globally (Hildreth et al., 2010; Fierstein et al., than predicted by models of progressive melt extraction 2012; Sruoga, 2015). This remarkable spatial and tem- (Charlier et al., 2008; Allan et al., 2013; Bindeman & poral concentration of rhyolite eruptions since the last Simakin, 2014; Wotzlaw et al., 2015). Thus, understand- glacial maximum, locally dated at c. 24 ka based on the ing the mechanisms of rhyolite genesis in a particular age of glaciated and unglaciated lava flows at LdM system can inform predictions of the processes and (Singer et al., 2000), has encircled the lake in the central timescales of the formation of a future, potentially large LdM basin and is unprecedented in the southern Andes eruptible silicic magma body. (Fig. 2; see also Table 1; Hildreth et al., 2010; Singer The importance of lower crustal differentiation in et al., 2014). Hildreth et al. (2010) presented several lines producing basalts and andesites in arc settings is well of evidence suggesting that these eruptions are derived recognized (e.g. Hildreth & Moorbath, 1988; Ownby from an integrated silicic magma system, most promin- et al., 2011); it has also been proposed that silicic ently: (1) rhyolite lavas erupted 10–12 km apart have magma is generated in the lower crust by partial nearly identical major and trace element compositions, Journal of Petrology, 2017, Vol. 58, No. 1 87
the last 1 5 Myr (Hildreth et al., 2010). During the same 100 km Tupungato period of time, frequent seismic swarms have occurred Santiago San Jose at similarly shallow depths near the Nieblas (rln) and Maipo- Barrancas (rcb) rhyolite flows, which are among the -34º Diamante Calabozos youngest in the volcanic field (Fig. 2; Singer et al., Caldera 2014). Initial gravity and magnetotelluric studies also Cerro Azul- Quizapu suggest the presence of a shallow, possibly growing, Talca Puelche magma system beneath the area of deformation at LdM Volcanic Field -36º 7.4 cm/yr Peru-Chile Trench (Singer et al., 2014; Miller et al., 2016). More recent geo- Laguna del Maule detic and geomorphological observations indicate that Concepción Tatara- the rate of uplift and inflation slowed slightly in 2013 (Le San Pedro Chile Domuyo Me´ vel et al., 2015) and that dozens of similar inflation Nevados de Chillan episodes have probably occurred throughout the -38º Antuco Holocene (Singer et al., 2015). Lonquimay Argentina The post-glacial eruptive chronology at LdM is cur- Llaima rently defined by only four 40Ar/39Ar ages obtained Villarrica nearly two decades ago (Singer et al., 2000) and the positions of lava flows relative to a paleoshoreline Mocho-Choschuenco -40º marking the highstand of the lake produced when the Puyehue-Cordón Caulle outlet gorge was dammed by the early rle rhyolite flow. Osorno Consequently, the age relations of eruptions occurring on opposite sides of the lake have been inferred based -74º -72º -70º -68º only on geomorphological features such as the extent of weathering and degree of pumice cover, hindering Fig. 1. Regional map of the SVZ between 33 and 41 S showing the interpretation of the temporal record. New 40Ar/39Ar the location of Laguna del Maule. Selected frontal arc volcanos 36 (triangles) and caldera systems and silicic volcanic centers and Cl surface exposure ages for late Pleistocene and (dark gray fields) are labeled for reference. The velocity of the post-glacial LdM lavas that refine the eruptive sequence Nazca plate relative to the South American plate is calculated are presented in this study. New whole-rock trace elem- using MORVEL (DeMets et al., 2010). ent compositions, Sr, Pb, and Th isotope ratios, and mineral thermobarometry are evaluated in the frame- suggesting that they are derived from a single homoge- work of this new geochronology to examine the tem- neous reservoir; (2) inclusions of mafic magma in rhyo- poral evolution of the rhyolite and rhyodacite magma dacite lavas are common, whereas mafic eruptions compositions. Models of magma evolution spanning have been rare and peripheral since the beginning of the last 150 kyr in the central LdM basin (earlier erup- post-glacial rhyolite volcanism, indicating that a broad, tions are sparse) are used to interrogate the continuity low-density magma body is blocking the ascent of and integration of the LdM magma system, the nature mafic magma. Consequently, the numerous post- and depth of the processes contributing to its evolution glacial silicic eruptions at LdM may represent a high through time, and the implications for the continuing temporal resolution sampling of the evolution of a volcanic unrest. large, shallow magma system. Several geophysical methods document continuing GEOLOGICAL SETTING volcanic unrest within the LdM basin that remains ac- The Quaternary LdM volcanic field is situated on the tive at the time of this writing. Geodetic data since 2007, crest of the Andes at 36 S in the Southern Volcanic obtained by continuous global positioning system Zone (SVZ) of central Chile (Fig. 1). Between 32 and (GPS) and interferometric synthetic aperture radar 37 S, the arc is characterized by a gradient in crustal (InSAR), record uplift at a rate in excess of 20 cm a–1, thickness from 35 km in the south to 60 km in the north among the fastest measured at a volcano not actively (Gilbert et al., 2006; Tassara et al., 2006; Tassara & erupting (Fournier et al., 2010; Feigl et al., 2014; Le Echaurren, 2012). This near doubling in thickness cor- Me´ vel et al., 2015). A model of an inflating sill at 5 km relates with a transition from dominantly basaltic an- depth produces the best fit of the measured deform- desite to amphibole-bearing intermediate products and ation pattern, with an estimated volume increase of well-documented gradients in trace element and radio- 3 –1 0 03–0 05 km a between 2007 and 2014 (Le Me´ vel genic isotope composition (Hildreth & Moorbath, 1988). et al., 2016). This probably transient rate is one to two Distinctively, the segment of the arc between 34 and orders of magnitude greater than the late Pleistocene to 37 S hosts several large Quaternary silicic volcanic cen- Holocene eruptive fluxes at the Southern Volcanic Zone ters in addition to LdM: the Maipo–Diamante Caldera (SVZ) frontal arc centers Mocho–Choshuenco and (Sruoga et al., 2012), the Calabozos Caldera (Hildreth Puyehue–Cordon Caulle (Singer et al., 2008; Rawson et al., 1984; Grunder & Mahood, 1988), Puelche et al., 2015) and the average eruptive flux at LdM over Volcanic Field (Hildreth et al., 1999), and Domuyo 88 Journal of Petrology, 2017, Vol. 58, No. 1
Rio Maule 1600 1800 1700 115 2200 230 1900 240 0 0 N 2100 Cajón Grande de Bobadilla 2828 3122 igcb 2500 Paso 2 2600 rle ig rdne 700 Pehuenche 2000 Cajón Chico de Bobadilla rdno 2874 rep bbc27 2800 rca rle 00 -36.0º igcb 115 bec 2600 igsp 25.7 ka 2500 2400 rdcn 2300 igcb 3.5 ka 19.0 ka 3080 L. Cari rdop rcn rcl Launa 0 00 2400 2680 rddm2300 0 rsl 303 rdsp 2500 0 2889 aam 0000 rsl 2600 2200 asp mpl 303129009000 2767 2800 3.3 ka apv Laguna 2900 3000 ram 2700 00 30 rdam 2600 00 2800 asm del 2500 29 2700 240 igsp apj Maule 2300 0 3175 rdcd 2162 2.1 ka 2883 anc apo rcd acn 2600 mnp aan dlp rdnp 8.0 ka 2300 0 rln 30000 2900 0 280228000 -36.1º rdep 20.0 ka mvc 2400 A. de la Calle 270027027 rdct 2500 2600 mcp 2500250 2600 2400 00 dlp mct 27 ras 2300 2600 2800 rdac 250 2855 rap 5.6 ka 0 22.42900 ka Aroyo de Palacios lveda 2300 3056 ú 2800 2400 2700 2600 2994 A. Sep 2500 1.9 ka rcb L. 2600 Negra 2162 700 Cajón de Troncoso 2 00 a Parva 2802808 l 90000 rcb-d 2 0 280800 Laguna 3037 Fea 30000 2888 2486 2800288000 rroyo de 270272700 rng A 2600 2600 270 rcb-d 00 0 25
00 CHILE 24 -36.2º Arroyo 2300 2200 2000 14.5 ka Puente de Tierra
ARGENTINA 2100 rcb-py Arr oyo Curamili
o 11.4 ka
-70.6º -70.5º -70.4º Central Laguna del Maule Volcanic Field Volcanic Vent Post Glacial Eruptions Pleistocene Eruptions Lava Flow Direction Rhyolite Rhyolite Pumice Rhyodacite Dacite/Rhyodacite Highway 115 Andesite Andesite International Boundary 05km Basalt/Mafic Andesite 3.5 Eruption age [ka] Contour interval 50 m Pleistocene ignimbrites Sample Locations all elevations masl igcb - 990 ka; igsp - 1.5 Ma Center of Deformation
Fig. 2. Simplified geological map of the central basin of the LdM volcanic field [after Hildreth et al. (2010)] showing sample loca- tions; unit names and abbreviations are listed in Table 1. Eruption ages are determined by 40Ar/39Ar except for the 36Cl age of unit rdcd; uncertainties associated with the 40Ar/39Ar ages are given in Table 2 and 36Cl data are given in the Supplementary Data. The center of uplift near the southwestern lake shore is an approximate location based on the InSAR model of Feigl et al. (2014). Journal of Petrology, 2017, Vol. 58, No. 1 89
Table 1: Laguna del Maule eruptive units mapped in Fig. 2
Abbreviation* Unit name* Eruption age† aam Andesite of Arroyo Los Mellicos 25 4 6 1 5ka acn Andesite of Crater Negro post-glacial anc Andesite north of Crater Negro post-glacial apj Younger andesite of West Peninsula 21 1 6 3 4ka apv Older andesite of West Penisula pre-glacial asm Andesite south of Arroyo Los Mellicos post-glacial asp Andesite of Laguna Sin Puerto <3 5ka bbc Basalt of Volcan Bobadilla Chica 153 6 7ka bec Basalt of El Candado 61 8 6 3 6ka dlp Dacite of Laguna del Piojo pre-glacial igcb Ignimbrite of Cajones de Bobadilla (rhyodacite) 990 6 13 ka igsp Ignimbrite of Laguna Sin Puerto (dacite) 1484 6 15 ka mcp Andesite of Crater 2657 post-glacial mct Andesite of Arroyo Cabeceras de Troncoso post-glacial mnp Andesite north of Estero Piojo post-glacial mpl Andesite of Volcan Puente de la Laguna 54 6 21 ka mvc Andesite of Volcan de la Calle 152 1 6 6 5ka ram Rhyolite of Arroyo Los Mellicos post-glacial; >19 ka rap Rhyolite of Arroyo de Palacios 22 4 6 2 0ka ras Rhyolite of Arroyo de Sepulveda 19–20 ka rca Rhyolite of Cajon Atravesado 710 6 13 ka rcb Rhyolite of Cerro Barrancas multiple flows; 11 4–1 9ka rcb-d Cerro Barrancas Dome Complex (rhyolite) 14 5 6 1 5ka rcb-py Cerro Barrancas Pyroclastic Flow (rhyolite) 11 4 6 1 1ka rcd Rhyolite of Colada Divisoria 2 1 6 1 3ka rcl Rhyolite of Cari Launa <3 3ka rcn Rhyolite of Cerro Negro 466 0 6 5 6ka rdac Rhyolite of Arroyo de la Calle 20 0 6 1 2ka rdam Rhyodacite of Arroyo Los Mellicos post-glacial; >19 ka rdcd Rhyodacite of Colada Dendriforme 8 0 6 0 8ka rdcn Rhyodacite of Northwest Coulee 3 5 6 2 3ka rdct Rhyodacite of Arroyo Cabeceras de Troncoso 202 6 41 ka rddm Rhyodacite of Domo del Maule 114 6 14 ka rdne Rhyodacite NE of Loma de Los Espejos post-glacial; >19 ka rdno Rhyodacite NW of Loma de Los Espejos post-glacial; >19 ka rdnp Rhyodacite north of Estero Piojo post-glacial rdop Rhyodacite west of Presa Laguna del Maule pre-glacial rdsp Rhyodacite of Laguna Sin Puerto <3 5ka rep Rhyolite east of Presa Laguna del Maule 25 7 6 1 2ka rle Rholite of Loma de Los Espejos 19 0 6 0 7ka rle-ig Espejos ignimbrite (rhyolite) post-glacial; >19 ka rln Rhyolite of Colada Las Nieblas Late Holocene rsl Rhyolite south of Laguna Cari Launa 3 3 6 1 2ka
*Abbreviations and unit names after Hildreth et al. (2010). †Ages are from Singer et al. (2000), Hildreth et al. (2010), Birsic (2015), and this study; all 40Ar/39Ar ages are calculated relative to the 1 1864 Ma Alder Creek Sanidine (Jicha et al., 2016).
Volcanic Complex (Miranda et al., 2006; Chiodini et al., calc-alkaline, medium- to high-K compositions typical 2014), each situated in the rear-arc relative to the basalt- of SVZ frontal arc volcanoes. Hildreth et al. (2010) found to andesite-dominated frontal arc volcanoes (Fig. 1). evidence for neither systematic variation in the slab sig- Owing to repeated glaciation and the remote, rugged nature across the volcanic field nor any significant con- terrain, it is not well appreciated that the productivity of tribution of back-arc, alkaline compositions. Basaltic Pliocene to Holocene silicic volcanism in this northern andesite to andesite dominates much of the preserved sector of the SVZ is comparable with that of the Andean eruptive history of LdM, but silicic (dacite–rhyolite) Central Volcanic Zone (Hildreth et al., 1984, 1999). eruptions have occurred throughout the volcanic field Hildreth et al. (2010) documented the most recent 1 5 during the Pliocene and Pleistocene (Hildreth et al., Myr of volcanic activity at LdM, which comprises more 2010). Two silicic ignimbrites are preserved in the LdM than 350 km3 of lava, tephra, and pyroclastic deposits lake basin (Fig. 2), the 1 5 Ma two-pyroxene dacite Sin ranging in composition from basalt to high-silica rhyo- Puerto Ignimbrite (igsp) and the 990 ka biotite rhyoda- lite erupted from at least 130 vents. The Quaternary cite Bobadilla Ignimbrite (igcb)(Birsic, 2015). Of these, eruptions overlie Paleogene to Neogene volcanic and only the Bobadilla caldera structure partially survived volcaniclastic rocks and Pliocene to Mesozoic plutons the subsequent glaciation and erosion. Two middle and sedimentary strata (Nelson et al., 1999; Hildreth Pleistocene rhyolitic lavas are preserved near the north- et al., 2010). LdM volcanic products are of tholeiitic to eastern shore of the lake, the 710 6 13 ka Rhyolite of 90 Journal of Petrology, 2017, Vol. 58, No. 1
Cajon Atravesado (rca) and the 466 0 6 5 6 ka Rhyolite Laguna Sin Puerto (rdsp) are crystal poor. The pheno- of Cerro Negro (rcn). The latter contains the most cryst assemblage is similar to that of the rhyolites but evolved compositions in the volcanic field (Hildreth all lack quartz and contain amphibole. Most rhyodacite et al., 2010). lavas contain fine-grained, partly glassy, basaltic andes- Singer et al. (2000) determined the timing of the last ite inclusions, frequently with quench textures, up to glacial retreat to be between 25 4 6 1 2 ka and 23 2 6 0 6 40 cm in diameter in the rhyodacites of Colada ka based on 40Ar/39Ar age determinations (recalculated Dendriforme (rdcd) and NW of Loma de Los Espejos to an Alder Creek Sanidine age of 1 1864 Ma; Jicha (rdno), but more commonly 1–10 cm in diameter. et al., 2016) for four eruptions, including one glaciated Similar inclusions are rare in the Rhyolite of Arroyo Los and three unglaciated lavas at approximately equal ele- Mellicos (ram) mini-dome but have not been found in vation in the LdM basin. This age is consistent with the any other rhyolite. moraine records east of the Andes between 47 and 46 S based on 3He, 10Be, and 26Al cosmogenic expos- 40 39 36 ure, 40Ar/39Ar, and 14C ages indicating that the last gla- NEW Ar/ Ar AND Cl AGES AND REVISED cial maximum occurred prior to 23 ka with deglaciation ERUPTION SEQUENCE well under way by 16 5ka(Kaplan et al., 2004; Hubbard An effort to document the LdM eruptive sequence based et al., 2005; Clark et al., 2009; Hein et al., 2010). The on the tephra stratigraphy and soil 14Cagesiscurrently post-glacial volcanism is concentrated in the LdM lake under way (Fierstein et al.,2012; Sruoga, 2015). basin, producing 36 silicic domes and coulees and doz- However, the construction of a 14C chronology at LdM is ens of explosive eruptions from at least 24 vents encir- challenging owing to a dearth of organic material. cling the lake (Fig. 2; Hildreth et al., 2010; Fierstein et al., Whereas 14C ages typically have lower uncertainties, 2012; Sruoga, 2015). Ten andesite flows emplaced since where suitable material is lacking, 40Ar/39Ar and 36Cl the glacial retreat, primarily along the western lake- ages offer alternative methods to date young volcanic shore, are of subordinate volume. Basaltic andesite is eruptions. Twenty-six 40Ar/39Ar incremental heating ex- rare since the most recent deglaciation and the young- periments, performed at the WiscAr Geochronology Lab est true basalt is the 61 8 6 3 6 ka basalt of El Candado (see Supplementary Data for details; supplementary (bec) erupted north of LdM (Fig. 2; Hildreth et al., 2010, data are available for downloading at http://www.pet recalculated to an Alder Creek Sanidine age of rology.oxfordjournals.org) yield plateau ages, all but one 1 1864 Ma; Jicha et al., 2016). containing more than 75% of the released 39Ar, and sup- Silicic eruptions at LdM were explosive and effusive port 12 eruption ages (Fig. 3; Table 2). We attempted to and generally of modest volume (<1 3km3; Hildreth determine 40Ar/39Ar ages for nearly all post-glacial lavas. et al., 2010; Fierstein et al., 2012). Continuing tephrostra- However, owing to their youth and high atmospheric Ar tigraphic investigations (Fierstein et al., 2012; Sruoga, contents, LdM products commonly yield small fractions 2015) both within the LdM basin and of distal deposits of radiogenic 40Ar (40Ar*). Micropumiceous rhyodacites in Argentina, are not discussed in detail here. However, and commonly vesiculated and glassy andesite flows of particular note, Fierstein et al. (2012) have identified a nearly all produced high 36Ar signals from which 40Ar* voluminous explosive eruption that produced flow and could not be resolved. Dense rhyolitic obsidian more fall deposits up to 6 m thick in Argentina 30 km south and east of LdM accounting for an order of magnitude commonly yields plateau ages; however, only approxi- greater volume than any single event mapped in the mately 50% of such samples produced resolvable ages. 39 central basin by Hildreth et al. (2010). This explosive Recoil of Ar during irradiation of volcanic glass can re- event pre-dates the rle lava flow that dammed the lake sult in spurious ages. This effect is mitigated for the LdM and thus is among the earliest post-glacial rhyolite lavas by a short irradiation time; age plateaux character- eruptions. However, its vent location and eruption age istic of recoil (i.e. decreasing apparent age with increas- remain uncertain. ing step heating temperature) are only sporadically Rhyolite flows preserved in the LdM basin are vitro- observed for sample aliquots subjected to longer dur- phyric and carry 5% modal phenocrysts; the rhyolite ation irradiation (see Supplementary Data). Several ex- of Arroyo Palacios (rap) and all but the latest of the periments display anomalously high ages in the low or Barrancas complex (rcb) flows are notably aphyric. high temperature steps. However, this behavior is con- Phenocrysts, when present, are dominantly plagioclase, sistent neither throughout the LdM sample suite, nor be- subordinate biotite, Fe–Ti oxide, sparse quartz, acces- tween aliquots prepared from single samples. The cause sory zircon, apatite, and very rare FeS inclusions in of these discordant steps is not clear, but they account magnetite; several rhyolites also contain scarce amphi- for less than 5% of the gas in single experiments and do bole. With the exception of the rhyodacite of Arroyo de not bias the reported ages. Inverse isochrons for all sam- la Calle (rdac) the rhyodacite lavas are concentrated in ples yield 36Ar/40Ar intercepts within uncertainty of the the western and northwestern basin. They are vitrophy- atmospheric ratio of Lee et al. (2006), indicating that ex- ric to micro-pumiceous and nearly all carry a pheno- cess Ar is not significant. The isochron and plateau ages cryst load of 10–25%, greater than any of the rhyolites; for each experiment are indistinguishable at 2r uncer- only the rhyodacites of the Northwest Coulee (rdcn) and tainty; thus the more precise plateau ages are preferred. Journal of Petrology, 2017, Vol. 58, No. 1 91
40 4.0 Southern Cari Launa Rhyolite (rsl) 30 3.3±1.2 ka 3.5 20 3.0±1.6 ka 3.7±2.1 ka 3.0 10 2.5 5.3±2.9 ka 0 40 39 Ar/ Ar0 = 293.8±8.6 n = 10 -10 2.0 0 0.2 0.4 0.6 0.8 1.0 0.0 0.4 0.8 1.2 1.6 2.0 2.4 50 Espejos Rhyolite (rle) 19.1±0.8 ka 5.0 40 39 19.0±0.7 ka Ar/ Ar0=296.4±2.6 40 n=18 4.0 3 30 17.7±2.5 ka 3.0
20 Ar x 10
40 2.0 Age [ka] Age 10 18.4±1.1 ka Ar/ 36 1.0 19.5±0.9 ka 0 0.0 0 0.2 0.4 0.6 0.8 1.0 0369 60 4.0 Rhyolite East of Presa 26.2±2.6 ka 40 39 50 Laguna del Maule (rep) 3.5 Ar/ Ar0 = 296.5±9.8 25.7±1.2 ka n = 13 40 3.0 30 2.5 20 2.0 25.8±1.3 ka 10 1.5 25.0±3.0 ka 0 1.0 0 0.2 0.4 0.6 0.8 1.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 Cumulative 39Ar Fraction 39Ar/40Ar
Fig. 3. Example 40Ar/39Ar age spectra and inverse isochrons for units rep, rle and rsl; values for all samples are available in the Supplementary Data. Plateau steps are colored boxes and ellipses; discordant, excluded steps are light gray. All uncertainties are- 6 2r and include the analytical and J uncertainties.
The eruption of the rle flow dammed the northern Early post-glacial eruptions outlet of the lake, causing the lake level to rise to The earliest of the recent silicic units erupted shortly 200 m above its modern level and cutting a prominent prior to deglaciation, forming the rhyolite east of the shoreline into all low-lying older rocks (Hildreth et al., Presa (dam) (rep)at25 7 6 1 2 ka in the northwestern 2010; Singer et al., 2015). To constrain better the dur- LdM basin. All subsequently erupted silicic units are ation of the lake highstand, we undertook 36Cl surface unglaciated, including the early voluminous pyroclastic exposure age determinations of the rdcd lava flow, event (Fierstein et al., 2012) and numerous andesite and which overruns the paleoshoreline is several places but rhyodacite flows and domes concentrated in the west- did not produce a resolvable 40Ar/39Ar age, and the ern and northwestern LdM basin. A single unglaciated shoreline itself where it is notched into the igcb ignim- andesite flow (apo) erupted in the south; this glassy, brite along the north shore of the lake (36Cl methods vesiculated lava did not produce resolvable 40Ar*. The and results are in the Supplementary Data). flow is largely buried by lake deposits and a pumice fan, The new age determinations are discussed in con- but apparently was erupted prior to the damming of the junction with the observations made during fieldwork lake. Rhyolite flows erupted on three sides of the lake in in support of the present work and by Hildreth et al. a relatively short time interval at the end of the EPG (2010) to improve the chronology of the post-glacial period: the Sepulveda rhyolite (ras) in the SE [which dir- eruptions. Whereas LdM erupted regularly following ectly overlies the 20 0 6 1 2 ka rhyodacite of Arroyo del the last glacial maximum, the rhyolitic volcanism is la Calle (rdac)], the 22 4 6 2 0 ka Palacios rhyolite (rap), clustered in two periods of high eruption frequency. and the 19 0 6 0 7 ka Espejos rhyolite (rle). An early post-glacial (EPG) group erupted prior to the damming of the outlet gorge at 19 ka. This was fol- Latest Pleistocene to Holocene eruptions lowed by a period of relative calm in much of the lake Volcanic activity waned throughout much of the LdM basin during the latest Pleistocene, with rhyolitic activ- basin following the end-EPG eruptions. The latest ity limited to the Barrancas complex in the SE basin. Pleistocene eruptions were restricted to the Barrancas Finally, silicic eruptions encircled the lake during the center (rcb) on the southeastern rim of the lake basin. Holocene (Fig. 4). An early episode of dome building is dated at 14 5 6 1 5 92 Journal of Petrology, 2017, Vol. 58, No. 1
Table 2: Summary of 40Ar/39Ar experiments
40 36 39 Sample no. K/Ca total Total fusion Ar/ Ari 6 2r MSWD Isochron n Ar % MSWD Plateau age [ka] 62r age [ka] 62r age [ka] 62r
Rhyolite of Cerro Barrancas, eat summit flow (rcb) AR-267 5 81 2 4 6 2 1 301 5 6 6 10 99 8 7 6 5 8 6 of 9 84 21 03 1 6 6 0 7 AR-267 5 80 4 4 6 3 1 323 6 33 0 24 -22 4 6 9 9 5 of 10 76 80 85 2 3 6 0 9 Combined isochron n ¼ 19: 301 0 6 5 71 00 1 3 6 8 0 Weighted mean n 5 2: 1 71 9 6 0 6 Rhyolite of Colada Divisoria (rcd) LdM-249* 5 06 7 7 6 3 2 305 2 6 8 10 12 -1 8 6 3 9 6 of 10 83 20 65 2 8 6 2 3 LdM-249* 5 18 3 2 6 2 7 300 7 6 5 10 19 -0 5 6 1 0 7 of 8 98 40 27 1 3 6 2 5 LdM-249 5 00 2 5 6 2 1 300 2 6 5 60 08 0 9 6 2 7 6 of 6 100 00 13 2 2 6 1 9 Combined isochron n ¼ 19: 301 1 6 3 40 28 0 3 6 0 3 Weighted mean n 5 3: 0 35 2 1 6 1 3 Rhyolite of South Cari Launa (rsl) ALDM-13-17 6 70 1 7 6 1 8 285 6 6 16 91 16 8 6 6 5 7 4 of 7 80 11 60 3 7 6 2 1 ALDM-13-17 6 60 3 0 6 1 6 298 2 6 11 40 25 3 2 6 4 1 6 of 8 96 20 20 3 0 6 1 6 Combined isochron n ¼ 10: 293 8 6 8 60 63 5 3 6 2 9 Weighted mean n 5 2: 0 65 3 3 6 1 2 Rhyodacite of the Northwest Coulee (rdcn) LdM-12-27 1 46 -1 2 6 2 4 294 7 6 8 41 54 5 7 6 4 5 5 of 7 89 7 1 38 3 5 6 2 3 Rhyolite of Cerro Barrancas, northern flow (rcb) LdM-210† 5 31 7 3 6 1 9 308 8 6 10 51 41 -2 2 6 2 4 5 of 6 98 22 43 5 2 6 2 7 LdM-210 5 25 3 9 6 2 4 298 7 6 16 22 95 2 6 6 2 9 5 of 6 97 92 18 2 7 6 3 1 LdM-210* 5 20 10 6 6 2 8 299 1 6 8 50 40 8 5 6 5 9 8 of 9 97 60 34 9 0 6 2 4 LdM-210* 5 26 8 2 6 3 2 301 5 6 12 70 90 3 1 6 2 7 7 of 9 86 10 77 4 9 6 3 0 LdM-210 5 12 5 6 6 1 4 331 0 6 37 90 63 1 9 6 1 9 7 of 7 100 01 14 5 7 6 1 2 Combined isochron n ¼ 27: 298 7 6 3 21 57 5 6 6 1 3 Weighted mean n 5 4: 1 47 5 6 6 1 1 Cerro Barrancas Pyroclastic Flow (rcb-py) CB-Curamilo A 6 18 12 6 6 4 9 318 6 8 60 47 -8 8 6 8 612of1584 11 211 5 61 3 CB-Curamilo A 5 88 10 0 6 5 1 294 9 6 5 81 3166 10 5 of 10 67 31 411 361 9 Combined isochron n ¼ 18: 298 7 6 4 81 211 4 6 7 1 Weighted mean n 5 2: 0 03 11 4 6 1 1 Cerro Barrancas Dome Complex (rcb-d) CB-2 4 90 13 7 6 1 6 297 4 6 2 70 80 15 8 6 3 6 9 of 9 100 0 0 80 14 5 6 1 5 Rhyolite of Loma de Los Espejos (rle) LdM-60 6 90 20 7 6 2 5 295 1 6 10 01 31 20 3 6 7 6 6 of 8 90 91 17 17 7 6 2 5 LdM-60 7 10 17 8 6 1 2 204 5 6 116 90 55 20 7 6 1 6 4 of 6 91 20 93 19 5 6 0 9 LdM-60 2 00 18 2 6 1 7 299 2 6 5 00 23 18 3 6 1 6 8 of 9 99 30 21 18 4 6 1 1 Combined isochron n ¼ 18: 296 4 6 2 60 73 19 1 6 0 8 Weighted mean n 5 3: 0 84 19 0 6 0 7 Rhyodacite of Arroyo de la Calle (rdac) LdM-213 2 05 20 9 6 1 8 294 1 6 9 70 59 22 6 6 3 3 7 of 7 100 00 62 21 2 6 1 5 LdM-213 2 03 19 9 6 1 8 286 1 6 12 00 46 22 4 6 3 7 5 of 7 93 91 34 18 8 6 1 8 Combined isochron n ¼ 13: 292 0 6 8 01 13 22 0 6 2 6 Weighted mean n 5 2: 1 28 20 0 6 1 2 Rhyolite of Arroyo Palacios (rap) LdM-12-23 6 20 22 4 6 2 0 293 4 6 12 90 47 23 6 6 3 6 7 of 7 100 0 0 49 22 4 6 2 0 Andesite of Arroyo Mellicos (aam) LdM-194 0 34 32 3 6 14 0 303 3 6 8 60 87 13 0 6 9 9 9 of 9 100 00 91 28 8 6 12 5 LdM-194 0 31 31 6 6 7 0 302 7 6 4 70 99 10 5 6 9 6 5 of 6 82 91 52 23 3 6 8 1 Combined isochron n ¼ 14: 303 0 6 4 10 76 10 8 6 6 3 Weighted mean n 5 2: 1 08 24 5 6 6 1 Rhyolite East of Presa Laguna del Maule (rep) LdM-12-32 7 50 24 7 6 3 4 291 9 6 28 00 82 26 3 6 5 8 8 of 8 100 00 73 25 0 6 3 0 LdM-12-32 7 00 27 0 6 1 3 296 6 6 10 60 42 26 3 6 2 8 5 of 6 83 40 35 25 8 6 1 3 Combined isochron n ¼ 13: 296 5 6 9 80 60 26 2 6 2 6 Weighted mean n 5 2: 0 56 25 7 6 1 2
Weighted mean plateau ages in bold are preferred; 2r uncertainties include the analytical and J uncertainties. *Monitored with the 28.201 Fish Canyon Sanidine (Kuiper et al., 2008); all other experiments were monitored with the 1.1864 Ma Alder Creek Sanidine (Jicha et al., 2016); †high MSWD; not included in weighted mean ka and is followed by an explosive event that produced rhyodacite flow, erupted near the western lake shore; pyroclastic flow deposits extending SE away from the these are the youngest units at sufficiently low elevation lake into Argentina (Fig. 2). A dense vitric clast from this within the lake basin to be subject to, but not affected pyroclastic deposit gave an age of 11 4 6 1 1 ka. These by, shoreline erosion. The youngest of the three north- earliest products of the Barrancas complex are exposed ern rcb flows yields an 40Ar/39Ar age of 5 6 6 1 1 ka; on its southern and eastern flanks and, therefore, are 40Ar* could not be resolved from either of the underly- not subject to shoreline erosion. Continued activity at ing flows. The rdcd flow yields a whole-rock 36Cl surface the Barrancas complex produced a series of rhyolite exposure age of 8 0 6 0 8 ka. The ages of the rdcd and flows, the northernmost of which, along with the rdcd northern rcb flows are consistent with a whole-rock 36Cl Journal of Petrology, 2017, Vol. 58, No. 1 93
(a) 5 km (c) rdne deglaciation rdno rcd lake rcl rle high stand rsl ram rep rdam aam asm apj apo acn rdnp East mnp anc apo rdep rcb summit aan rln ras rdac rcb rap
rng rcb-py Early rcb-d Post glacial ras 25.7 - 19.0 ka rdac rap (b) rdep South rdcd rdnp mnp asp asm rdcn rdsp rcl apj acn rsl anc aan rdcd West rcd rdsp rln asp rdcn mcp mct rle ram rdam rcb-d rcb rdne rdop Latest Pleistocene rng aam rep to Holocene North 14.5 - ≤ 1.9 ka rcb-d rcb-py 30 25 20 15 10 5 0 rcb-py Age [ka]
Fig. 4. Post-glacial eruptive sequence of central LdM basin lavas. Fill colors are the same as in Fig. 2. (a) The distribution of EPG eruptions—those erupted prior to and including the rle flow that dammed the outlet gorge producing the highstand of the lake. (b) The distribution of latest Pleistocene to Holocene eruptions. (c) The relative eruptive sequence constrained by 40Ar/39Ar ages from Singer et al. (2000) and this study; the timing of the drawdown of the lake highstand is constrained by a 9 5 6 0 1ka36Cl surface ex- posure age of the highstand shoreline cut into igcb tuff. Black outlined boxes are 40Ar/39Ar and 36Cl ages, with the width corres- ponding to the 2r uncertainty. Gray outlined boxes are inferred eruption age ranges based on field relationships; the widths are set relative to the nominal ages of the constraining events. surface exposure age of 9 5 6 0 1 ka for the shoreline Outside the Holocene south–SE rhyolite focus, the cut into igcb in the northern lake basin. rhyodacite of the Northwest Coulee (rdcn) erupted from ThemiddletolateHolocene saw rhyolite eruptions a vent near the crest of the NW basin wall and extends from four centers in the southern and eastern lake nearly down to the current lake level 350 m below. This basin (Fig. 4). A significant explosive eruption from the prominent flow is dated at 3 5 6 2 3 ka and is mantled Cari Launa complex (Fierstein et al., 2012)wasfol- by the andesitic cinder ring of Laguna Sin Puerto (asp), lowed by the older of two Cari Launa rhyolite flows which was subsequently intruded by the rhyodacite of (rsl)at3 3 6 1 2ka,theRhyoliteofColadaDivisoria Sin Puerto (rdsp). These eruptions likewise emanated (rcd)at2 1 6 1 3 ka, and the small rcb floweastofthe from a vent on the crest of the NW basin wall. Two Barrancassummitat1 9 6 0 6 ka. Neither the upper- small andesitic fissure eruptions, the andesite of Crater most western rcb flow nor the rhyolite of Colada 2657 (mcp) and the andesite of Arroyo Cabeceras de Las Nieblas (rln) produced resolvable 40Ar*, but on the Troncoso (mct), occurred 6 km west of the SW lake- basis of their similar lack of pumice cover and shore. The ages of these eruptions are not well con- uneroded morphology, they are probably of compar- strained; however, mcp scoria blankets the post-glacial able age to the rcd and eastern summit rcb lavas and rhyodacite south of Estero Piojo (rdep) mini-domes to thus are among the most recent eruptions in the vol- the north, but not the mct craters, indicating that both canic field. are younger than rdep and, although they are at a 94 Journal of Petrology, 2017, Vol. 58, No. 1
30 1000 (a) (d) T-SP 25 Central LdM basin 800 Greater LdM 20 Pleistocene ignimbrites 600 15 Sr [ppm]
Th [ppm] Th 400 10
5 200
0 0 500 1000 (b) (e) T-SP 400 800 T-SP 300 600 K/Rb
Zr [ppm] Zr 200 400
100 200
0 0 15 25 (c) (f)
12 20
9 T-SP 15 Rb/Y 6 La/Yb 10
3 5
0 4550 55 60 65 70 75 80 4550 5560 65 70 75 80
SiO2 [wt. %] SiO2 [wt. %]
Fig. 5. Variation of selected trace elements with SiO2 for lava and pumice erupted in the central LdM basin during approximately the last 150 kyr. Data for the 1 5 Myr history of the entire volcanic field, including the Pleistocene igcb and igsp ignimbrites (Hildreth et al., 2010; Birsic, 2015) and T-SP (Dungan et al., 2001) are plotted for comparison. The typical 2r uncertainties associated with the central LdM data are smaller than the symbols. The central LdM data show less dispersed ranges and trends relative to the larger LdM volcanic field and T-SP. The REE and Y ratios of the igcb and igsp ignimbrites notably diverge from those of the post-glacial si- licic lavas. Plots of major element variation are available in the Supplementary Data. higher elevation than the high strandline, possibly post- Whereas many major and trace elements, such as date the rle eruption as well (Hildreth et al., 2010). K2O, MgO, Th, U, Rb, and Pb, evolve monotonically with increasing SiO2, several display prominent inflec- tions in variation diagrams (Fig. 5 and supplemen- WHOLE-ROCK GEOCHEMICAL RESULTS tary figures). Between 52 and 60% SiO2, high field Major and trace elements strength elements (HFSE) (except Ti), large ion litho- Lavas erupted during the last 150 kyr in central LdM phile elements (LILE) (except Sr), light REE (LREE), range from basalt to high-silica rhyolite. Primitive lavas, heavy REE (HREE), and Y increase with increasing SiO2. rare throughout the SVZ, are absent from central LdM Between 60 and 68% SiO2, Zr and LREE level off and as indicated by the modest Mg# ( 53) and low K/Rb TiO2, MREE, Y, and P2O5 begin to decrease. Ba concen- ratios (369–242) of the basalt and mafic andesite sam- trations increase to 65% SiO2 but vary little in the more ples. The major and trace element evolution of central evolved lavas. Between 68 and 70% SiO2, Zr concentra- LdM generally mirrors that of the entire 1 5 Myr erup- tions begin to decrease and the depletion of Sr with tive history of the larger volcanic field (Hildreth et al., increasing SiO2 becomes greater. 2010) and the frontal arc Tatara–San Pedro complex (T– SP; Dungan et al., 2001). Central LdM trace element Sr and Pb isotope ratios compositions form narrow arrays in elemental variation The Sr and Pb isotope compositions of the central LdM plots compared with the range observed in the volcanic units, measured at the University of Wisconsin–Madison field as a whole (Fig. 5). The Pleistocene LdM ignim- ICP–TIMS Isotope Laboratory [Sr by thermal ionization brites igcb and igsp are notably enriched in rare earth mass spectrometry (TIMS) and Pb by multicollector in- elements (REE), particularly middle REE (MREE), Y, and ductively coupled plasma mass spectrometry (MC-ICP- Zr compared with the post-glacial silicic lavas. MS); see the Supplementary Data for details], display Journal of Petrology, 2017, Vol. 58, No. 1 95
15.75 the SVZ (Nelson et al., 1999). No lava with a comparably Mz intrusions radiogenic Sr isotope ratio has erupted in the central LdM since the middle Pleistocene. The range of central
Pz basement LdM is also similar to, but slightly narrower than, those 15.65 found at the nearby Pleistocene silicic centers including Pb sed the Puelche Volcanic Field (0 70386–0 70440; Hildreth 204 Pε-N intrusions et al., 1999) and the Loma Seca Tuff and associated Pb/ 207 15.55 Quaternary frontal arc lavas (0 70380–0 70433; Grunder, 1987). MORB SA-N NHRL Th isotopes The Th isotopic compositions, measured by MC-ICP-MS 15.45 at the University of Wisconsin–Madison ICP–TIMS 18.2 18.3 18.4 18.5 18.6 18.7 18.8 Isotope Laboratory (see Supplementary Data for de- 206Pb/204Pb tails), span a narrow range with modest disequilibrium Fig. 6. The 206Pb/204Pb and 207Pb/204Pb ratios of central LdM in both U- and Th-excess (Fig. 8; Table 4). The age- basin lavas (red squares); data are given in Table 3. Also corrected (230Th/232Th) activity ratios of the LdM lavas shown is the Northern Hemisphere Reference Line (NHRL; 0 Hart, 1984), the composition of South Atlantic N-MORB (SA- range from 0 773 to 0 808, among the lowest yet meas- NMORB; Douglass et al., 1999), Mesozoic (Mz) and Paleogene ured in the SVZ. The rhyolites and rhyodacites display a to Neogene (Pe–N) intrusive rocks (Lucassen et al., 2004), modest U-excess, up to 5%, and a narrow range of Paleozoic (Pz) intrusive and metamorphic basement (Lucassen 230 232 et al., 2004), SVZ sediments (Hildreth & Moorbath, 1988; ( Th/ Th)0 ratios, 0 793–0 808. The mafic lavas show Lucassen et al., 2010; Jacques et al., 2013), and Quaternary a greater diversity of Th isotopic compositions. The SVZ frontal arc lavas (Davidson et al., 1987; Gerlach et al., 230 232 ( Th/ Th)0 ratios of mafic lavas are nearly all lower 1988; Hildreth & Moorbath, 1988; Hickey-Vargas et al., 1989; McMillan et al., 1989; Jacques et al., 2013; Holm et al., 2014). and have a 50% larger range, 0 773–0 800, than those of The LdM lavas yield a narrow range of 207Pb/204Pb isotopic the silicic eruptions. Most are in 2–5% Th-excess. ratios compared with the frontal arc edifices and are distinct Quenched mafic inclusions hosted in units rdno and from those of the Paleozoic to Mesozoic basement, indicating rdne and the basaltic andesite lava mpl are in 3–4% that any assimilation was of younger, more primitive crust. 230 232 U-excess and have low ( Th/ Th)0 ratios spanning a similar range to the Th-excess lavas (Fig. 8). limited variation. Ratios of 87Sr/86Sr range from 0 70407 to 0 70422, 206Pb/204Pb from 18 615 to 18 646, 207Pb/204Pb from 15 606 to 15 622, and 208Pb/204Pb from 38 521 to THERMOMETRY AND BAROMETRY 38 565 (Fig. 6; Table 3; Supplementary Data Figs A6 and Two-oxide thermometry 207 204 A7). Whereas the Pb/ Pb ratio does not vary coher- The compositions of Fe–Ti oxides were determined by ently with major or trace element composition, higher electron microprobe at the University of Wisconsin– 87 86 206 204 208 204 Sr/ Sr, Pb/ Pb, and Pb/ Pb ratios are corre- Madison (see the Supplementary Data for details). lated with increasing SiO2 (Supplementary Data Fig. A7). In the LdM rhyolites and rhyodacites, the ulvo¨ spinel The late Pleistocene to early post-glacial andesites apj content of magnetite ranges from Ulv13 to Ulv25 and and aam have elevated 87Sr/86Sr ratios similar to those hematite content of ilmenite from Hm25 to Hm31. The of the silicic eruptions, but slightly less radiogenic average Fe–Ti oxide compositions of the apj andesite 206Pb/204Pb and 208Pb/204Pb ratios compared with the flow are Ulv50 and Hm15. Oxides in both the rhyolites more mafic units. In contrast, quenched mafic inclusions and rhyodacites span the compositional range in the northern rhyodacite domes rdno and rdne have observed in the silicic units; however, the highest 87 86 Sr/ Sr ratios similar to those of the basalts and mafic ulvo¨ spinel contents found in rhyolites are limited to the 206 204 208 204 andesite lavas, but higher Pb/ Pb and Pb/ Pb products of the Cari Launa (rcl, rsl) center. 87 86 ratios. The Sr/ Sr ratio of the modest-volume andesite Fe–Ti oxide temperatures calculated using the cali- scoria eruption asp is similar to that of apj, aam,andthe bration of Ghiorso & Evans (2008) are 760–850 C for the silicic eruptions, but also has the most radiogenic rhyolites, 796–854 C for the post-glacial rhyodacites, 206Pb/204Pb and 208Pb/204Pb ratios of this sample suite. and 760 C for the late Pleistocene rhyodacite rddm 87 86 The range of the central LdM Sr/ Sr ratios is not- (Fig. 9). Silicic units yield an fO2 1 19–1 32 log units ably narrow compared with regional volcanic centers above the Ni–NiO buffer (NNO). Oxides from the (Fig. 7). The LdM volcanic field as a whole has a wider Younger Andesite of the Western Peninsula (apj) gave a 87 86 range of Sr/ Sr ratios of 0 70388–0 70435 and one temperature of 1017 C and fO2 0 3 log units greater high outlying ratio, 0 70483, from the 430 ka rhyolite of than NNO. The range of temperatures produced for Cerro Negro (rcn; Hildreth et al., 2010). The Miocene multiple oxide pairs from each sample is 35 C for all Risco Bayo–Huemul plutonic complex exposed beneath but three samples, commensurate with the 630 C un- the Tatara San Pedro volcanic complex contains volu- certainty typically ascribed to the two-oxide thermom- metrically minor domains with 87Sr/86Sr ratios signifi- eter (Ghiorso & Evans, 2008). The later erupted Cari cantly greater (>0 7050) than those of juvenile lavas in Launa rhyolites and unit rdcd produced temperature 96 Journal of Petrology, 2017, Vol. 58, No. 1
Table 3: Whole-rock Sr and Pb isotopic compositions
Sample Unit 87Sr/86Sr 2SE 206Pb/204Pb 2SE % 207Pb/204Pb 2SE % 208Pb/204Pb 2SE % n
LDM-12-25 aam 0 70419 0 00001 18 618 0 00005 15 613 0 00005 38 532 0 00004 2 LDM-12-19 apj 0 70419 0 00001 18 623 0 00004 15 612 0 00004 38 540 0 00004 2 ALDM-13-09 asp 0 70419 0 00001 18 648 0 00004 15 614 0 00004 38 570 0 00004 2 LDM-12-34 bec 0 70412 0 00001 18 623 0 00006 15 611 0 00006 38 533 0 00004 1 LDM-12-31 mnp 0 70409 0 00001 18 622 0 00005 15 613 0 00005 38 538 0 00004 1 LDM-12-15 mpl 0 70408 0 00001 18 623 0 00008 15 621 0 00007 38 550 0 00004 1 LDM-12-23 rap 0 70418 0 00001 18 638 0 00005 15 614 0 00006 38 557 0 00004 2 LDM-13-13 rcb 0 70419 0 00001 18 636 0 00004 15 615 0 00004 38 557 0 00004 2 ALDM-13-14 rcb 0 70420 0 00001 18 633 0 00005 15 612 0 00005 38 549 0 00004 2 LDM-12-07 rcd 0 70422 0 00001 18 632 0 00006 15 612 0 00006 38 549 0 00004 2 LDM-12-08 rcl 0 70419 0 00001 18 634 0 00005 15 613 0 00005 38 552 0 00004 1 LDM-12-11 rdac 0 70420 0 00001 18 638 0 00005 15 614 0 00005 38 553 0 00004 1 LDM-12-17 rdcd 0 70418 0 00001 18 640 0 00004 15 613 0 00005 38 557 0 00004 3 LDM-12-17i rdcdi 0 70410 0 00001 18 621 0 00006 15 614 0 00007 38 534 0 00004 1 LDM-12-27 rdcn 0 70413 0 00001 18 630 0 00004 15 612 0 00005 38 538 0 00004 2 ALDM-13-10 rddm 0 70420 0 00001 18 633 0 00006 15 613 0 00005 38 547 0 00004 1 LDM-12-03 rdne 0 70421 0 00001 18 636 0 00006 15 616 0 00005 38 551 0 00004 1 ALDM-13-01 rdnei 0 70407 0 00001 18 630 0 00010 15 606 0 00011 38 535 0 00004 1 LDM-12-33 rdno 0 70420 0 00001 18 635 0 00003 15 613 0 00003 38 551 0 00004 2 LDM-12-33i rdnoi 0 70412 0 00001 18 637 0 00003 15 617 0 00003 38 561 0 00004 1 LDM-12-16 rdnp 0 70411 0 00001 18 631 0 00006 15 612 0 00007 38 546 0 00004 1 ALDM-13-08 rdsp 0 70414 0 00001 18 636 0 00004 15 614 0 00005 38 537 0 00004 1 LDM-12-32 rep 0 70420 0 00001 18 637 0 00004 15 612 0 00005 38 550 0 00004 1 LDM-12-04 rle 0 70419 0 00001 18 637 0 00004 15 613 0 00004 38 550 0 00004 3 LDM-12-30b rle p 0 70420 0 00001 18 637 0 00004 15 613 0 00004 38 554 0 00004 1 LDM-12-21 rln 0 70420 0 00001 18 636 0 00005 15 615 0 00005 38 560 0 00004 1 ALDM-13-17 rsl 0 70419 0 00001 18 634 0 00004 15 613 0 00004 38 554 0 00004 2 Standard analyses 2SD 2SD 2SD 2SD NIST SRM-987 0 71028 0 00001 30 NBS-981 16 940 0 004 15 496 0 004 36 720 0 011 22 NBS-982 36 754 0 026 17 161 0 006 36 749 0 015 11 BCR-2 0 70505 0 00002 18 756 0 029 15 628 0 009 38 720 0 100 5 (Sr), 7(Pb) AGV-2 0 70402 0 00002 18 861 0 041 15 619 0 005 38 535 0 044 6 (Sr and Pb)
87Sr/86Sr ratios are reported as measured; age correction is inconsequential for these young samples.
Central LdM SVZ 36º S spreads greater than 60 C. The younger Cari Launa lava Basin flow (rcl) and associated pumice cone produced a simi- Rhyolite Greater LdM lar range of temperatures that in aggregate is 812– Rhyodacite Loma Seca Tuff 884 C; the lowest temperature in this range is more Basalt - Andesite Tatara - San Pedro Mafic Inclusions than 2r from the mean. Excluding this temperature nar- Puelche Volcanic Field rows the range to 845–884 C. Unit rdcd produced a 0.7046 similarly wide range of 823–889 C; all calculated tem- peratures are within two standard deviations of the mean. 0.7044 Sr
Amphibole thermobarometry 86 0.7042
Amphibole and plagioclase crystals in five rhyodacite Sr/ 87 lavas (rdac, rdne, rdno, rdcd, and rdcn) were analyzed 0.7040 by electron microprobe at the University of Wisconsin– Madison. The plagioclase compositions are utilized to 0.7038 estimate the magma water content required for the amphibole barometer calibration of Putirka (2016);a 0.7036 0 200 400 600 800 more thorough interrogation of the plagioclase com- Sr [ppm] positions will be the subject of a future contribution. The anorthite content of plagioclase rims ranges from Fig. 7. Comparison of the central LdM basin 87Sr/86Sr as a func- An to An . Using the hygrometer of Waters & Lange tion of Sr content with those of nearby volcanic centers includ- 19 43 ing T-SP (Davidson et al., 1987), the rear-arc Puelche volcanic (2015), the plagioclase rim and rhyodacite whole-rock field (Hildreth et al., 1999), the Calabozos Caldera complex– compositions yield a mean water content for each unit Loma Seca Tuff (Grunder, 1987), and older eruptions through- ranging from 4 5to5 0 wt % at 850 C and 250 MPa; a out the LdM volcanic field (Hildreth et al., 2010). The regional grand mean of 4 8 wt % is adopted for the amphibole data are plotted age corrected; the age correction is insignifi- cant for the central LdM lavas and these ratios are plotted as calculations. The Waters & Lange (2015) hygrometer re- measured values. The central LdM lavas show a notably nar- quires an estimate of the crystallization pressure, but is row range compared with these nearby systems. Journal of Petrology, 2017, Vol. 58, No. 1 97
Laguna del Maule Puyehue - Cordon Caulle Llaima (a) (b) 1.0 Quizapu 0.850 Osorno and small Puyehue centers Mafic lava 33º - 41º S historic mafic eruptions Mafic inclusion Rhyodacite 0.825 Rhyolite o 0.9 Rhyolite Glass Th)
232 0.800 Th/ 230 ( 0.8 0.775
equiline equiline 0.7 0.750 0.70.91.11.30.70 0.75 0.80 0.85 0.90 (238U/232Th) (238U/232Th)
Fig. 8. Equiline plots of age-corrected Th isotope activity ratios for central LdM lavas and pumice erupted in the last 150 kyr. (a) The LdM data compared with those measured at other SVZ volcanic systems (Hickey-Vargas et al., 2002; Sigmarsson et al., 2002; Jicha 230 232 et al., 2007; Reubi et al., 2011; Ruprecht & Cooper, 2012). Central LdM lavas have among the lowest ( Th/ Th)0 activity ratios yet measured in the SVZ. (b) Detail equiline plot of the LdM data including the Th-excess mafic lavas and U-excess silicic products and 230 232 rhyodacite-hosted mafic enclaves. The uncertainties in the ( Th/ Th)0 data include those of the ages used to correct the meas- ured ratios for decay since eruption. Dashed tie-lines connect mafic inclusions to their host rhyodacite. relatively insensitive to this parameter. Over a range of distinct differentiation pathways involving diverse crus- 100–900 MPa, the calculated water content varies by tal assimilants and crystallizing assemblages. In the fol- only 0 15 wt %. Thus, the inclusion of a pressure esti- lowing sections we explore the following: (1) the mate in the hygrometry calculation does not bias the processes that have contributed to the geochemical amphibole barometry. characteristics of the LdM lavas, particularly the sources The LdM amphiboles are pargasite to magnesio- of U- and Th-excess; (2) whether these processes devi- hornblende based on the classification scheme of ate significantly from those inferred at frontal arc volca- Hawthorne et al. (2012). Amphibole formulae based on 23 noes; (3) the processes promoting the more oxygen atoms, pressures, and temperatures are calcu- homogeneous isotopic compositions of the rhyolites lated using the method of Putirka (2016). The equilibrium compared with the mafic samples; (4) the temporal co- melt SiO2 is calculated to assess equilibrium with the host herence of the thermo-chemical evolution of the LdM magma; amphiboles that deviate by more than 4 wt % magma system; (5) the implications for the structure from the host composition, the uncertainty associated and state of the modern magma reservoir. with the equilibrium SiO2 estimate, are not included in the pressure calculations (Putirka, 2016). The resulting dataset Crustal contributions to mafic magmas comprises 12–38 amphibole analyses for each unit and Frontal arc centers in the central and southern SVZ yields average crystallization pressures of 190–250 MPa commonly show relatively narrow ranges of radiogenic with uncertainties of 30–50 MPa (Fig. 9). These pressures isotope ratios, despite trace element evidence for sig- are consistent with those calculated by the less precise, nificant crustal interaction, owing to limited isotopic but magma composition-independent, barometer calibra- contrast between the primary mafic magmas and the ju- tions of Ridolfi et al. (2010) and Ridolfi & Renzulli (2012). venile crust (e.g. Davidson et al., 1987; Dungan et al., Pressure- and magma composition-independent amphi- 2001). Uranium-series isotopes are a sensitive tracer of bole thermometry produces a range of 828–933 C, which magma evolution in arc systems as they provide infor- overlaps the two oxide temperatures from the rhyodacite mation about the nature of mantle and crustal compo- lavas, but also extends to higher temperatures. nents, the processes leading to their mixing, and in some cases the timescales of these processes (e.g. Hickey-Vargas et al., 2002; Turner et al., 2003, 2010; DISCUSSION Jicha et al., 2007, 2009; Reubi et al., 2011; Ankney et al., The narrow compositional arrays of the central LdM 2013). Mafic lavas in U-excess are common in arc set- basin lavas suggest a common magmatic origin tings and are often attributed to the flux of slab fluids to (Hildreth et al., 2010). However, divergent correlations the mantle wedge (e.g. Turner et al., 2003). Less com- among radiogenic isotope ratios and inflections in the mon Th-excess continental arc magmas are generally trajectory of trace element variation diagrams suggest thought to reflect a garnet signature inherited from the 98 Journal of Petrology, 2017, Vol. 58, No. 1
Table 4: Whole-rock and glass 230Th–238U compositions
238 232 230 232 230 232 230 238 Sample Unit Age (ka) Th (ppm) U (ppm) ( U/ Th) 2SE ( Th/ Th) 2SE ( Th/ Th)0 2SE* ( Th/ U)0 n
LDM-12-25 aam 25 4 6 1 59 20 2 27 0 748 0 004 0 781 0 005 0 790 0 008 1 057 1 LDM-12-19 apj 21 1 6 3 48 04 2 03 0 765 0 005 0 778 0 005 0 781 0 007 1 021 1 ALDM-13-09 asp <3 57 94 2 07 0 792 0 005 0 790 0 005 0 790 0 005 0 998 2 LDM-12-34 bec 61 8 6 3 63 09 0 77 0 754 0 005 0 765 0 005 0 773 0 012 1 025 1 LDM-12-15 mnp <24 4 06 1 09 0 814 0 005 0 800 0 005 0 798 0 008 0 981 1 LDM-12-31 mpl 54 6 21 6 76 1 65 0 742 0 004 0 770 0 005 0 788 0 020 1 062 1 LDM-12-23 rap 22 4 6 2 022 05 5 95 0 819 0 005 0 800 0 005 0 795 0 007 0 970 1 LDM-13-13 rcb <318 80 5 05 0 815 0 005 0 799 0 005 0 798 0 005 0 980 1 ALDM-13-14 rcb 14 5–5 620 97 5 61 0 812 0 005 0 799 0 005 0 798 0 006 0 982 1 LDM-12-07 rcd 2 2 6 1 220 59 5 49 0 810 0 005 0 798 0 005 0 798 0 005 0 986 2 LDM-12-08 rcl <3 320 14 5 41 0 815 0 005 0 799 0 005 0 798 0 005 0 979 1 LDM-12-11 rdac 20 0 6 1 219 99 5 42 0 822 0 005 0 798 0 005 0 793 0 007 0 964 1 LDM-12-17 rdcd 8 00 6 0 84 19 60 5 39 0 834 0 005 0 798 0 005 0 797 0 005 0 956 2 LDM-12-17i rdcd i 8 00 6 0 84 3 39 0 88 0 784 0 005 0 798 0 005 0 798 0 005 1 017 1 LDM-12-27 rdcn 3 5 6 2 315 20 4 12 0 822 0 005 0 799 0 005 0 798 0 006 0 971 1 ALDM-13-10 rddm 114 6 14 18 22 4 94 0 823 0 005 0 815 0 005 0 802 0 027 0 974 1 LDM-12-03 rdne 25 7–19 016 36 4 36 0 809 0 005 0 799 0 005 0 797 0 007 0 985 1 ALDM-13-01 rdne i 25 7–19 05 39 1 42 0 798 0 005 0 778 0 005 0 773 0 008 0 968 1 LDM-12-33i rdno i 25 7–19 06 16 1 67 0 823 0 005 0 783 0 005 0 774 0 008 0 941 1 LDM-12-16 rdnp <24 15 34 6 18 0 828 0 005 0 802 0 005 0 799 0 009 0 965 1 ALDM-13-08 rdsp <3 516 97 4 61 0 824 0 005 0 804 0 005 0 804 0 005 0 976 1 LDM-12-32 rep 25 7 6 1 223 42 6 32 0 819 0 005 0 802 0 005 0 797 0 008 0 973 1 LDM-12-04 rle 19 0 6 0 723 50 6 32 0 816 0 005 0 803 0 005 0 800 0 007 0 980 2 LDM-12-30b rle pum 19 0 6 0 723 25 6 20 0 810 0 005 0 808 0 005 0 808 0 007 0 998 1 LDM-12-21 rln <319 08 5 14 0 817 0 005 0 800 0 005 0 800 0 005 0 980 1 ALDM-13-17 rsl 3 3 6 1 220 89 5 56 0 808 0 005 0 797 0 005 0 796 0 005 0 985 2 LDM-12-07 rcd glass 2 1 6 1 221 76 5 81 0 811 0 005 0 794 0 005 0 794 0 005 0 979 1 LDM-12-04 rle glass 19 0 6 0 723 67 6 36 0 815 0 005 0 803 0 005 0 801 0 007 0 982 1 LDM-12-21 rln glass <320 25 5 42 0 812 0 005 0 798 0 005 0 798 0 005 0 982 1 ALDM-13-14 rcb glass 14 5–5 621 16 5 66 0 812 0 005 0 794 0 005 0 793 0 006 0 977 1 LDM-12-08 rcl glass <3 520 78 5 60 0 817 0 005 0 801 0 005 0 801 0 005 0 980 1 Standard analyses 2SD 2SD BCR-2 5 87 1 68 0 869 0 002 0 876 0 006 6 AGV-2 6 01 1 86 0 938 0 002 0 946 0 005 5
230 232 *( Th/ Th)0 uncertainty includes that of the eruption age. mantle or lower crust owing to its affinity for U over Th estimated as the average of that of Palme & O’Neill
(DU/DTh ¼ 2 3–12 9; e.g. Rubatto & Hermann, 2007; Qian (2003), will yield Th-excess similar to that measured in & Hermann, 2013). In the SVZ, correlations among the LdM lavas (Fig. 10). However, these low extents of fluid-mobile trace elements, 10Be/9Be, and U-excess in melting favor silica-undersaturated melts inconsistent frontal arc basalts have been interpreted as a slab fluid with the silica-saturated to -oversaturated lavas erupted control of the primary Th isotope signature (Hickey- at LdM. Thus, the Th excess at LdM most probably re- Vargas et al., 2002; Sigmarsson et al., 2002). flects a greater extent of mantle melting and a contribu- However, subsequent U-series studies of several SVZ tion from garnet-bearing crust (GBC). The 207Pb/204Pb centers, including LdM, call into question the ubiquity of ratios of the LdM lavas are distinct from those of the this relationship. The enrichment of fluid-mobile elem- more radiogenic Paleozoic to Mesozoic basement, indi- ents in the SVZ is modest compared with volcanic arcs cating that this crustal component must be relatively globally (e.g. Ba/Th < 300) and is only weakly correlated primitive (Fig. 6; Luccassen et al., 2004). with U-excess (Fig. 10; Supplementary Data Fig. A8). Models of lower crust melting are calculated using ex- Moreover, correlations between fluid-mobile element en- perimental phase equilibria and partition coefficients richment and U-excess can result from crustal assimila- from the literature (see the Supplementary Data for tion rather than variations in the slab fluid signature model parameters). The composition of the lower crust (Reubi et al.,2011). Whereas the addition of slab fluids to is estimated using the global average of Rudnick & Gao 230 232 the mantle wedge plays an important role in promoting (2003); the narrow range of the ( Th/ Th)0 ratios of U-excess at some frontal arc centers, several mechan- the LdM lavas suggest that the initial U/Th ratio of the isms could contribute to their decoupling in the SVZ: (1) crustal component is similar to that observed at LdM, long magma residence (>350 kyr) following the addition and thus the estimated crustal composition is adjusted of the fluid component to the mantle wedge allows the accordingly. Batch melting of GBC (e.g. Berlo et al.,2004; U-excess signature to decay away (Hickey-Vargas et al., Hora et al.,2009) and the formation of garnet during de- 2002); (2) the addition of a Th-enriched sediment melt to hydration melting of initially garnet-free amphibolite the mantle wedge would mitigate the fluid-derived U en- (Wolf & Wyllie, 1993; Ankney et al.,2013)havebeenpro- richment (Jacques et al.,2013). posed to explain Th-excess in continental arc settings. In the absence of significant fluid-derived U enrich- The latter, although appropriate for the large Th-excess ment, 3–6% partial melting of garnet lherzolite mantle observed in Cascade lavas (Jicha et al.,2009; Ankney (e.g. Ottonello et al., 1984), with a composition et al.,2013; Wende et al.,2015), yields large Th-excess Journal of Petrology, 2017, Vol. 58, No. 1 99
-10 0 Rhyodacite lavas (a) (b) rdcn -11 rdcd LdM rdac Loma Seca 100 rdne -12 Rhyodacite Tuff Holocene rdno LdM NNO 2 -13 Rhyolite VTTS EPG O f 200
Log -14 post-Oruanui P [MPa]
-15 2 Bishop Tuff + 300 QFM -16
Glass Mt. -17 400 650 700 750 800 850 900 800 850 900 950 T [ºC] T [ºC]
Fig. 9. Results of mineral thermobarometry for central LdM eruptions. (a) T–fO2 plot for central LdM silicic eruptions. Fields show the range of temperatures and oxygen fugacities for the Loma Seca Tuff (Grunder & Mahood, 1988), Bishop Tuff (Hildreth & Wilson, 2007), Glass Mountain rhyolites (Metz & Mahood, 1991), post-Oruanui rhyolites (Sutton et al., 2000) and the Valley of Ten Thousand Smokes rhyolites (VTTS; Hildreth, 1983). Reference T–fO2 curves for the nickel–nickel oxide buffer (NNO) and 2 log units above the quartz–fayalite–magnetite buffer (QFM þ 2) are shown, illustrating the highly consistent T–fO2 buffering of the LdM erup- tions. (b) Temperatures and pressures derived from amphibole compositions for LdM rhyodacite lavas. The pressure calculation assumes a magma with 4 8wt%H2O based on plagioclase hygrometry (Waters & Lange, 2015). Each point is a single spot analysis and has uncertainties of 630 C and 6160 MPa (Putirka, 2016). The bars on the left of the plot are the average pressure and associ- ated uncertainty for each unit. The pressures of the Holocene lavas are nominally 50–60 MPa less than, but within uncertainty of, those of the EPG units. and HREE depletions inappropriate for the SVZ between mantle-derived melts and the continental (Supplementary Data Fig. A3). Mixing of 10% partial crust. Moreover, these processes vary little from those melts of garnet-bearing crust and mantle reasonably re- inferred at frontal arc centers throughout the SVZ (e.g. produces the range of Th-excess and REE compositions Davidson et al., 1987; Hildreth & Moorbath, 1988; found at LdM; however, the presence of U-excess mafic McMillan et al., 1989; Dungan et al., 2001; Costa & lavas requires an additional explanation (Figs 10 and 11). Singer, 2002; Jicha et al., 2007). Thus, whereas the con- LdM mafic lavas in U-excess could be interpreted as centration of rhyolite at LdM is exceptional, the underly- reflecting the slab fluid signature only partially over- ing mafic magmatic processes are not. printed in the lower crust. However, these samples are enriched in incompatible elements relative to the basalts Shallow vs deep origin of rhyolite and mafic andesites in Th-excess, indicating that the The LdM silicic lavas are depleted in Ti, P, Sr, and Y, U-excess mafic lavas have experienced greater inter- have negative Eu anomalies, and have lower Dy/Yb action with a crustal component. In contrast to garnet ratios relative to the andesites (Fig. 5; Supplementary production by amphibolite dehydration, the formation of Data Fig. A5). These trends indicate a shift in the differ- clinopyroxene during the melting of plagioclase- and entiation regime from that of mafic magmas primarily amphibole-bearing crust (garnet-free crust; GFC) can influenced by assimilation of crustal melts. Annen et al. produce U-excess (Fig. 11; Beard & Lofgren, 1991; Berlo (2006) suggested that the majority of compositional di- et al.,2004). Holocene intermediate lavas at T-SP were versity of volcanic rocks is imparted by lower crustal produced, in part, by the melting of hornblende-bearing processes. This model is inconsistent with the relatively mafic intrusions similar to T-SP xenoliths (Costa & shallow crystallization pressures determined by amphi- Singer, 2002). A 10% dehydration melt of this material bole barometry at LdM. However, it is possible that the yields 6% U-excess, commensurate with the range observed in the LdM lavas (Fig. 10). amphibole is late crystallized and does not capture the Mixing among the mantle, GBC, and GFC end- high-pressure differentiation history of the silicic mag- members, each produced by 10% partial melting, can mas. Differences in phase equilibria and the compos- explain the Th isotope and trace element diversity of ition of potentially assimilated rocks between the deep the LdM mafic lavas. Variation of the Th isotope ratios and shallow crust would impart predictable, divergent with the Zr/Th and La/Yb ratios forms offset arrays with geochemical trends during the generation of silicic the largest Th-excess, found in units mpl and aam, magma that are compared with the LdM compositions associated with higher La/Yb and lower Zr/Th ratios. to judge the plausibility of differentiation in the lower vs This offset is consistent with variable mixing, 5–30%, of upper crust. the GBC and mantle melts. Additional mixing with a We utilize Rhyolite-MELTS (Gualda et al., 2012) 10% partial melt of GFC yields the range of U-series dis- to simulate fractional crystallization of an andesitic equilibrium observed in the LdM mafic lavas (Fig. 11). LdM parental magma at a range of pressures Thus, despite a relatively limited range in isotopic com- (150–1050 MPa), initial water contents (1–6 wt %), and positions, the LdM lavas reflect extensive interactions fO2 buffers (QFM to QFM þ 2, where QFM is quartz– 100 Journal of Petrology, 2017, Vol. 58, No. 1
350 water content. High pressures, water contents and (a) Quizapu Villarica SEC reducing conditions promote the early stabilization of LdM Puyehue 300 pyroxene at the expense of plagioclase and magnetite, equiline Llaima Puyehue SEC Villarica Osorno producing large depletions in MgO over a narrow range 250 of SiO2, inconsistent with the LdM compositions (Fig. 12). Moreover, Gaulda & Ghiorso (2013) argued that the 200 increasing stability of quartz with depth precludes the generation of rhyolite by high-pressure fractional Ba/Th 150 crystallization. 100 MELTS is not well calibrated for hydrous intermedi- ate to silicic compositions saturated in amphibole. 50 However, in this case, the SiO2/MgO ratio of LdM amphibole (2 6–3 4) is between those of orthopyroxene (2–3) and clinopyroxene (3 3–4 4) predicted by MELTS 0.6 0.8 1.0 1.2 1.4 1.6 1.8 such that the crystallization of either two pyroxenes or (238U/ 230Th) 0 amphibole would have a similar impact on the magma 0.84 SiO2/MgO ratio. Whereas some model misfit may result (b) from the prediction of pyroxene rather than amphibole
quiline e crystallization, the agreement between the MELTS mod- 0.82 eling and amphibole barometry indicates that the sup- LdM silicic pression of plagioclase and magnetite crystallization is 0 Garnet 2 3 4610 15 lavas 0.80 lherzolite the more important factor. Thus, MELTS simulations of Th) melting hydrous systems must be interpreted with caution, but 232 30 can yield useful first-order phase equilibrium con- Th/ 0.78 5101520 1510 5 230 straints even when amphibole is present. ( Garnet-bearing crust melting Garnet-free The physical plausibility of a viscous rhyolite magma 0.76 crust melting ascending >30 km through the crust is questionable (e.g. Rubin, 1995). Even if it were possible, the similarity of the rhyolite 87Sr/86Sr ratios to those of the mafic 0.74 0.65 0.70 0.75 0.80 0.85 0.90 and rhyodacite lavas (Fig. 7) weigh against a deep crust (238U/232Th) origin. Following differentiation in the lower crust, Sr- depleted rhyolite would then traverse the crustal col- Fig. 10. Sources of U-series disequilibrium in central LdM lavas. (a) Plot of SVZ U-series disequilibrium data for mafic umn that includes highly radiogenic Paleozoic to lavas compared with the Ba/Th ratio, an indicator of fluid en- Mesozoic rocks (Lucassen et al., 2004; Supplementary richment. Volcanic centers, including small eruptive centers Data Fig. A6). The inevitable assimilation of even small (SEC) associated with larger edifices, are listed in the legend in geographical order from north (Quizapu) to south (Osorno) amounts (<5%) of this material would produce higher 87 86 along the arc (Fig. 1). Some centers display evidence of cou- and more variable Sr/ Sr ratios in the rhyolites than pling between fluid enrichment and U-excess; however, this observed. The more radiogenic 87Sr/86Sr ratios, correlation may also result from crustal overprinting of the >0 7046, of the mid-Pleistocene rcn rhyolite erupted in slab signature (Reubi et al., 2011). The range of Ba/Th in the Th-excess lavas is similar to that in U-excess and thus a strong the eastern LdM basin and the most-evolved domains coupling between fluid enrichment and U-series disequilibrium of the Miocene plutonic complex beneath T-SP (Nelson is not evident in the SVZ. Data sources: Sun (2001), Hickey- et al., 1999; Hildreth et al., 2010) potentially reflect as- Vargas et al. (2002), Jicha et al. (2007), Reubi et al. (2011) and Ruprecht & Cooper (2012). (b) The U-series disequilibrium ex- similation of this material; however, the modestly radio- pected during melting of the garnet-bearing mantle, garnet- genic, homogeneous 87Sr/86Sr ratios of the post-glacial bearing lower crust, and garnet-free crust (see the rhyolites do not. Taken together, the isotope ratios of si- Supplementary Data for model parameters). The Th-excess ap- parent in the mafic LdM samples (red squares) can be pro- licic LdM lavas, the incongruity between the predicted duced by melting with residual garnet in either the mantle or phase equilibrium and the LdM major element compos- lower crust. U-excess in several mafic andesites and the silicic itions, and shallow crystallization pressures recorded lavas reflects the overprinting of the garnet signature by partial by amphibole barometry rule out generation of the LdM melting of garnet-free crust rather than U enrichment imparted by a subduction fluid (see text). rhyolites in the lower crust.
Shallow hybridization and fractional fayalite–magnetite) to evaluate the conditions in which crystallization the LdM rhyolite magma formed. Each model is cooled The narrow range of Th isotope ratios and uniform from the calculated liquidus to c. 700 C, depending on U-excess of the silicic lavas contrast with the more var- model convergence at low melt fractions. The variation ied mafic compositions (Fig. 8). Fractionating Th from U of SiO2 and MgO of the LdM lavas is best reproduced in the upper crust to produce the silicic compositions by shallow, oxidizing conditions and a moderate initial from a parental melt in Th-excess is not Journal of Petrology, 2017, Vol. 58, No. 1 101
1.15 (a) (b)
1.10 10% GFC melt 10% GFC 50 50 melt 1.05 30 0 LdM 30 LdM silicic 50 silicic
Th) lavas 0 10 lavas 5 0 equiline 10 equiline 230 1.00 3 30
U/ 10% mantle 238 ( 10 melt 10 10% 0.95 10% GBC 5 mantle 5 10 melt 10 melt 10% 20 20 GBC 50 30 30 50 0.90 melt
0.85 10 20 30 40 50 0102030 Zr/Th La/Yb
(c)Mantle melt + 30% GBC melt (d) Mantle melt + 5% GBC melt
100
10 % GFC melt sample/chondrite 10 30 50 1 La Ce Nd Sm Eu Dy Yb La Ce Nd SmEu Dy Yb
Fig. 11. A mixing model to explain the variation of U-series disequilibrium and the trace element composition of the mafic LdM lavas. The mixing endmembers are 10% melts of garnet lherzolite mantle, garnet-bearing crust (GBC), and garnet-free crust (GFC) (see the text and Supplementary Data). (a) and (b) show the variation of U-series disequilibrium with the Zr/Th and La/Yb ratios pro- duced by first mixing mantle and GBC melts. Subsequent mixing with a 10% melt of garnet-free crust produces the range of Th- and U-excess observed in the LdM mafic samples (red squares). The offset arrays of LdM data are consistent with varying mixing proportions of the mantle and GBC end-members. (c) and (d) show chondrite-normalized (Sun & McDonough, 1989) REE patterns produced by 10%, 30%, and 50% mixing of the GFC endmember with a melt composed of 5% or 30% mixing of GBC with the man- tle melt, compared with those of the mafic LdM lavas (gray field). straightforward. Crystallization of major phases will not several volcanoes in the Andes, Cascades, and Alaska significantly increase the U/Th ratio, but accessory (Garrison et al., 2006; Jicha et al., 2007; Turner et al., phases such as apatite, titanite, allanite, and monazite 2010; Ankney et al., 2013). This transition has variously have greater leverage (Berlo et al., 2004). Of these, only been ascribed to mixing with a U-excess endmember apatite is common at LdM. Rare, possibly xenocrystic, derived from small degrees of partial melting with re- titanite has been recovered by heavy liquid separation sidual accessory phases, hydrothermal alteration of from the large, early tephra eruption, but not from any assimilated wallrock, and variation in the contribution other LdM rhyolite; neither allanite nor monazite are of a subduction component through time. The garnet- present. The crystallization of sufficient apatite or titan- free crustal component evident at LdM offers an alter- ite to produce U-excess from a Th-excess mafic magma native explanation. The requirement of garnet in the is not consistent with the P2O5 and MREE compositions production of Th-excess limits this process to the lower- of the LdM lavas: fractionation of 0 3% titanite most crust. Thus, only rapidly ascending magmas
(DTh ¼ 18 7, DU ¼ 7, DDy ¼ 935, DYb ¼ 393; Bachmann would preserve a garnet-derived Th isotope signature. et al., 2005)or3 2% apatite (DTh ¼ 2 82, DU ¼ 1 9; Those that stall in the middle to upper crust and further Condomines, 1997) is required to produce the observed differentiate will have greater opportunity to interact change in the U/Th ratio. The crystallization of these with GFC and acquire U-excess. Amphibole, common in phases in this quantity would decrease the Dy/Yb ratio arc crust, is produced both by direct crystallization and by a factor of seven and the P2O5 composition by 1 4wt by reaction between clinopyroxene and ascending hy- %, respectively; both are approximately four times drous melt. Costa et al. (2002) advocated the latter greater than the variation observed in the central LdM mechanism for the generation of amphibole beneath lava compositions. Thus, the crystallization of accessory T-SP and it also probably occurs at LdM. The subse- phases cannot account for the U-excess observed in the quent melting of amphibole-bearing crust has been pro- silicic lavas. posed as an important source of melt and volatiles in The eruption of mafic magma in Th-excess and volcanic arcs more generally (e.g. Davidson et al., 2007, evolved magma in U-excess has been observed at 2013); thus, the production of clinopyroxene during 102 Journal of Petrology, 2017, Vol. 58, No. 1 amphibole dehydration may be an under-appreciated 4 Model Conditions source of U-excess in intermediate to evolved continen- 210 MPa, 3% H2O, QFM+2 tal arc magmas. 210 MPa, 3% H2O, QFM The evolution of the major and many trace element 3 210 MPa, 5% H O, QFM+2 compositions from the andesitic to silicic magmas is con- 2 600 MPa, 3% H O, QFM+2 sistent with the fractionation of the plagioclase þ 2 amphibole þ biotite þ Fe–Ti oxide þ apatite 6 zircon as- 2 semblage observed in the rhyodacite and rhyolite lavas.
The saturation of zircon yields prominent inflections in the %] [wt. MgO evolution of the Zr concentration (Fig. 5); deviations from 1 the expected closed-system evolution would favor more extensive open-system processes. Zr and Th are similarly (a) incompatible in major phases and thus, prior to zircon sat- 0 uration, fractional crystallization would produce compar- 55 60 65 70 75 80 able enrichments in both elements. In central LdM, the SiO2 [wt. %] 1.0 modest difference in the Zr concentrations of the rhyoda- mt cite and andesite lavas is incongruent with the two-fold 0.9 cpx difference in the Th concentrations. 0.8 We first consider a model of zircon-free fractional 0.7 crystallization of an andesite parental magma utilizing a plag range of Zr partition coefficients, the anhydrous mineral 0.6 opx assemblage predicted by the best-fit MELTS model (Fig. 0.5 12), and a hydrous mineral assemblage in which amphi- melt phase fraction 0.4 bole crystallizes in place of pyroxene (Table 5). None of bt+ 0.3 these fractional crystallization pathways are able to pro- qtz duce the variation in Zr composition of the intermediate 0.2 LdM lavas (Fig. 13). The zircon saturation temperature 0.1 210 MPa, 3% H2O, QFM+2 (b) of most of the post-glacial rhyodacites is less than but 0 within uncertainty of the two-oxide temperature, indi- 1050 950 850 750 cating they may have been zircon saturated—based on 1.0 opx the zircon saturation model of Watson & Harrison 0.9 cpx (1983); none are zircon saturated using the model of 0.8 Boehnke et al. (2013). Thus, the Zr contents of the rho- 0.7 dacite lavas could be produced by fractional crystalliza- plag tion including a small but increasing modal per cent 0.6 mt zircon or could reflect open-system processes. 0.5 The two-oxide temperature of the andesite apj is 0.4 gt+